Content uploaded by Albi Salm
Author content
All content in this area was uploaded by Albi Salm on Apr 21, 2019
Content may be subject to copyright.
post-publication comment on:
Dive Risk Factors, Gas Bubble Formation, and Decompression Illness in
Recreational SCUBA Diving: Analysis of DAN Europe DSL Data Base
https://doi.org/10.3389/fpsyg.2017.01587
This work of european DAN is not only intellectualy stimulating but as well of paramount importance to the further
development of safety for recreational divers. It may also help to mitigate the somewhat heated and superfluous
debate in the (technical) diving community as to which decompression model (perfusion, diffusion, dual phase) or
which dive computer might be the best.
Nevertheless there are a couple of boundary conditions which will probably evade the non-diving reader.
1a) The 83 : 17 relation of participants, which yields a ratio of ca. 4,9 of males over females is an
imbalance which probably may not reflect properly the european recreational divers population. This might be a
first indicator of a biased database.
1b) As well the „mean“ age seems to reflect a somewhat non-standard diver population; the majority of divers
being usually younger. As per nearly all human activities, there is a drop-out rate: for ca. one
instructor/TEC/advanced diver there are approx. 100 freshman. So here the beginner and intermediate diver
population seems to miss.
1a & 1b could be checked against the statistics of the issued diver certifications of the major training agencies like
PADI and/or SSI.
2) 320 cases of DCS would yield:
an average rate of ca. 64 p.a.
and an overall rate of 320 / 39.099 of ca. 0,81 % which, both, are substantially more than reported
elsewhere (ca. 1: 10.000 in [1], p.544; and as well in [2], p. 151).
Certainly, the divisor is not known and the rates are taken as a surrogate. And, as well, other, more elaborate
studies, would reveal, for eg.: Decompression illness in divers treated in Auckland, New Zealand, 1996-
2012. (Ref.: https://www.ncbi.nlm.nih.gov/pubmed/24687481): that there are 520 DCI cases in 17 years, which
would equate to ca. 31 p.a. and thus yield a factor of ca. 2 lower.
Similarily a recent study showed (Ref.: Svendsen Juhl C, Hedetoft M, Bidstrup D, Jansen EC, Hyldegaard O.
Decompression illness treated in Denmark 1999–2013. Diving and Hyperbaric Medicine. 2016 June;46(2):87-91.)
205 DCI cases over 15 years; reduced to recreational cases, the rate is ca. 11 p.a. The factor male/female ca.
3.8. This study reveals that ca. 80 % of the DCI victims are up to an intermediate certification level, and ca. 50 %
consider themselves as with relatively little experience.
Another example, albeit taken from the military diving community (DJRS, the Dive Jump Reporting System of
USN (United States Navy); Ref.: http://divingresearch.scripts.mit.edu/militarydivingdata/) which sports with with 4
cases of DCS and 5 of missed decompression out of 768,851 dives, collected from 2008 to 2015.
This might be as well a 2nd. indicator of a biased database.
3) 39,099 dives per 2,629 divers would yield ca. 15 dives per diver. It might be somewhat speculative, but
15 dives out of an average diving carreer do probably not represent an average sample concerning
both dive depth an dive time. Thus this could imply that the uploaded samples have been the most recent ones or
the most spectacular ones (with a topical tendency from October 2017 of 63.248 from 5.326 divers, giving an
average of 12, sinking). Especially when considering that approx. 15 to 20 individuals, each donating in the
average ca. 500 dives, thus contributing a substantial amount of profiles and thus decreasing the average of the
rest.
4) Studies with volunteer participants regularly imply often a self-selected group, the sample usually not reflecting
the real population. This could imply another, i.e. the 3rd. bias: participants which like to deal with the technical
pecularities of transferring log-files from a dive computer to a PC, then converting the stored dive log-files to a
DAN compatible format, then uploading these files from the PC to the DAN DSG portal and finally filling in the ca.
20 statements per uploaded dive.
5) The average dive depth / dive time given as mean ± standard deviation, would imply, at least for a somewhat
naive reader, sort of normal distributions for these variables, which would be, in my personal experience,
relatively unprobable. Otherwise the study fails to reveal the statistical connection between diver B doing a dive in
country X to depth z with a buddy-pair C, diving in country Y to depth to 0,5 * z with diver A (me) contributing in
country DE a controlled dive in a decompression chamber to depth 2 * z, the mean being clearly z, but obviously
of only limited intellectual value (Ref.: https://www.divetable.info/skripte/50m_deco.pdf. The funny side of things is
that these profiles normally earn a big yellow smiley in the DAN DSG portal, thus warning of an already medium
DCS risk).
For eg. the above cited USN study reveals clearly skewness, with a slope (note the log scale!) from shallow to
deep and more probably of a certain Poisson type than Gaussian. Thus, a frequency analysis in appropriate
classes (depth bins with 5 or 10 m resp. for the dive time) would have given a clearer picture.
6) As well the mean of 27.1 m (range 5–104), where the .1 is clearly a statistical artefact, which could have been
dropped happily: dive computers tend to give the first digit not precisely. Anecdotal reports [3], [4] and controlled
laboratory experiments [5], [6] indicate this very clearly. Additionaly, a lot of dive computer manufacturers fail to
demonstrate a proper temperature drift compensation for their products.
The value of 27 could be as well an indicator of a certain bias: the suspected missing of beginner and
intermediate dive-profiles, being in the 6 to 18 m range for beginners and in the 15 to 30 m range for intermediate
divers. Even more so, when considering beginner and intermediate divers as relatively neutral to decompression-
theories, -calculation and -tables; some of them not even owning a dive computer.
7) To exclude Trimix makes sense, probably there is another mechanism of bubble arterialisation and other inert
gas kinetics due to Helium (for the non-diving reader: Trimix is a breathing gas, consisting in various fractions of
the 3 gases Oxygen, Helium and Nitrogen (thus Tri-), whereas simple compressed Air or Nitrox are not.) As well
the ZH-L framework from Buehlmann et al. [7], used in this study, has been, up to now, not really challenged with
trimix for multi-level diving. (And b.t.w. this algorithm is diverging around a compartment half-time of 1,005 min,
which, used unmodified, would render it useless for the intended analysis of breathhold diving).
8) Dives, for eg. to 104 m depth on air, will yield profiles of an extreme spike form which are probably not in line
with common diver behaviour, be it recreational, military or commercial: the limitation of breathing gas supply
makes bottom times very short, especially when dived with a single tank; which the study tacitly implies. If done
otherwise, the study should reveal it.
9) Also, for the non-diving reader: a dive on Air in the 3-digit range is subjected to inert gas narcosis, which is
likely to start beyond 40 m and oxygen toxicity, beyond 80 m, which makes these profiles, operationally wise, not
only relatively dangerous, but, to put it mildly, somewhat „experimental“: the ascent and descent rates are not in
line with standard procedures. Thus one could question the statistical wisdom of not excluding these experimental
dives.
10) In conclusion, the study leaves open a couple of questions resp. room for improvement:
Diver biometrics and dive circumstances are entered through the divers themselves. How is the quality
of these inputs assessed?
Screening for PFO or other individual susceptibilities?
Blinding of operators, recieving the doppler signals? Control group?
Table 3 reveals a conundrum of multi-collinearity: how is this adressed?
11) Nobody should be caught by surprise, that the mapping of a deterministic perfusion model (ZH-L) to a
stochastic phenomenon (DCS) is of only limited success. Thus the relative failure of printed decompression tables
or dive computers. Once again, as per remark # 7, the ZH-L (or, basically all perfusion models) was never really
challenged with extended multi-level or reversed dive profiles, common in recreational diving. The described
modifications ([7], p. 157, 196) to allow for real-time calculation being marginal: the clear message is that the M-
values (or, in ZH-L parlance, the a- & b coefficients) derived from box-profiles, and, maybe, the spectrum of
compartment half-times need repair.
This even more so, when considering that dive computers are „black boxes“ for the diver: leaving the user
completely in the dark, how a decompression algorithm is implemented and which constants are used. Thus it is
also of no surprise that for a given box-profile the calculated stop times for the decompression stops differ easily
with a factor of 4 to 8; even if the manufacturers in question claim to have implemented a „real ZH-L“ (Ref.:
https://www.divetable.info/skripte/HBO-RMT.pdf and [10]).
One of the really important findings is, that the dives seem to be basically in the „safe zone“: thus nobody should
be caught by surprise, that the group of „medium“ compartments is involved. The mentioned slower ascent rate,
although not specified in the study, and „deep“ or deeper and longer stops, give the fast compartments time to de-
saturate while the medium and slow ones still saturate. For real world recreational diving, the take home message
seems to be:
„If you go slow, go even slower!“ (especially in the shallow 9 to 6 m zone).
12) Now, finally in taking points # 1) to 4) as a basis, chances are that there could be a bias of the actual
database; one of the confounding factors being diver experience and another one the liking of handling purely
technical problems. My private speculation and personal experience is, that this relatively special group of highly
trained and motivated (mediterranean) divers, which dedicates a lot of their spare time for the DAN DSG portal
tends to dive in a way that is, in some way or another, disjunct with the population of recreational divers, thus
prone to a higher rate of DCS.
I would not go so far as with Altman, who states: „Misuse of statistics is unethical, as well it is shoddy science.“
[8]; but clearly a couple of tenthousand non-DCS dives with moderate time/depth profiles have to be added. As
well a functional peer-review process, in-line with established statistical thinking would be a benefit.
References:
[1] Bennett and Elliott's Physiology and Medicine of Diving, Alf Brubakk, Neuman et al., 5 th Ed. Saunders,
ISBN 0-7020-2571-2
[2] Bove and Davis' DIVING MEDICINE, Alfred A. Bove, 4 th. edition, Saunders 2004,
ISBN 0-7216-9424-1
[3] https://www.divetable.info/skripte/Ox_Tox.pdf
[4] https://www.divetable.info/skripte/G2_OX_TOX.pdf
[5] UHMS: ASM 2012, Session F118; NOT ALL ARE CREATED EQUAL – OPERATIONAL VARIABILITY IN 49
MODELS OF DIVING COMPUTER. Azzopardi E, Sayer MDJ, UK National Facility for Scientific Diving, Scottish
Association for Marine Science, Dunstaffnage Marine Laboratories, Dunbeg, Oban, Argyll, Scotland
[6] Elaine Azzopardi and Martin Sayer (2012) Estimation of depth and temperatures in 47 models of diving
decompression computer. International Journal of the Society for Underwater Technology, Vol. 31, No. 1, pp 3 -
12
[7] Tauchmedizin, Albert A. Bühlmann, Ernst B. Völlm (Mitarbeiter), P. Nussberger; 5. Auflage in 2002, Springer,
ISBN 3-540-42979-4
[8] Douglas G. Altman: Misuse of statistics is unethical, BMJ, 281, 1. Nov 1980, p. 1183
[9] Edmonds, Carl; Bennett, Michael et al.(2016) Diving and Subaquatic Medicine, Fifth Edition, CRC Press; ISBN
978-1-4822-6012-0
[10] Salm, A. (2012) Variations in the TTS: where do they come from? International Journal of the Society for
Underwater Technology, November 2012. SUT, Vol. 31, No. 1, pp. 43 - 47, 2012
